{"id":"ec64bdfa-a28b-4307-9b16-d8dec40fb887","arxiv_id":"1908.02241","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proves generalized integration by parts formulas for Bessel bridges with arbitrary boundary values and constructs a weak gradient dynamics for the two-dimensional Bessel bridge.","lead":"This paper extends a family of integration by parts formulas to Bessel bridges and Bessel processes with arbitrary starting and ending points, and it constructs a Markov evolution for the two-dimensional Bessel bridge. The results provide new analytic tools for the conjectural Bessel stochastic partial differential equations, which are linked to random interface and pinning models.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 4.1 is not proved: the existence, locality, and quasi-regularity of the closure of (E², FC_b^∞(K)), and the identity (4.4), are deferred to [23, Thm 5.1.3] and [7, Prop 5.1]. Corollary 4.2 and Theorem 4.7 collapse if this fails.","rationale":"The reader identifies Proposition 4.1 as the weakest assumption, and after reading the manuscript I agree. The Section 3 integration by parts formulae are supported by explicit computations from the Laplace transforms of squared Bessel bridges, and the bridge case is obtained by conditioning on X₁ and identifying Laplace transforms of continuous functions. The main unresolved point is therefore not the algebra of Section 3 but the construction of the Dirichlet form and the associated Markov process in Section 4. The paper itself acknowledges the omission: the proof of Proposition 4.1 is deferred to Theorem 5.1.3 of [23] and Proposition 5.1 of [7], with the arguments attributed to Rongchan Zhu and Xiangchan Zhu. Since the existence of the reversible Markov process is exactly what Theorem 4.7 supplies, the omitted proof is load-bearing. Corollary 4.2 and the weak SPDE identification in Theorem 4.7 would collapse if quasi-regularity or the identity (4.4) failed. I do not see an internal inconsistency in the IbPF derivations, and I have not identified a counterexample; the concern is of the omitted-proof type. Therefore the reader's CONDITIONAL verdict is appropriate and I would not change it.","tokens_in":32890,"tokens_out":30536,"duration_ms":307131,"concrete_test":"Supply the full argument for Proposition 4.1, following Theorem 5.1.3 of [23] and Proposition 5.1 of [7] line by line for the specific map j₂(z) = ‖z‖ and the measure ν₂ = µ₂ ∘ j₂^{-1}. In particular, prove closability by checking E²(f,g) = Λ²(f∘j₂,g∘j₂) on FC_b^∞(K) and then extending by approximation; prove locality by showing the energy vanishes on functions with disjoint support; prove quasi-regularity by exhibiting a compact nest and a separating algebra for the closure. Then verify that Corollary 4.6 and Theorem 4.7 follow from this proposition together with Lemma 4.5, without importing additional unproved steps from [7]. If the proof cannot be completed as stated, the dynamical theorem should be downgraded to conditional on an external argument.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing condition for the dynamical half of the paper is Proposition 4.1. It asserts that the gradient form (E², FC_b^∞(K)) is closable, that its closure is a local quasi-regular Dirichlet form on L²(ν₂), and that the isometric identification E²(f,g) = Λ²(f∘j₂,g∘j₂) holds on the closure. No proof is given: the text says it can be proved as in Theorem 5.1.3 of [23] or Proposition 5.1 of [7], and that the arguments were communicated by R. Zhu and X. Zhu. Corollary 4.2, which provides the Markov diffusion process with continuous paths, and Theorem 4.7, which identifies the weak SPDE dynamics via a martingale plus a zero-energy additive functional, both depend on this proposition. The projected IbPF identity in Corollary 4.6, and hence the identification of the drift term, additionally depends on the projection operator Π whose existence uses the same quasi-regular closure. This is an omitted proof of existence of the object that the main dynamical theorem is about, not a minor technical convenience. The Section 3 IbPFs are independent of this assumption and appear to be argued in full, so the conditional status is concentrated in Section 4.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends the integration-by-parts formulae (IbPFs) of Elad Altman and Zambotti, originally proved for δ-dimensional Bessel bridges from 0 to 0, to Bessel bridges on [0,1] with arbitrary boundary values a,a'≥0 and to Bessel processes with arbitrary initial conditions. The IbPFs are stated in Theorem 3.1 for all δ>0, with special formulas for δ=1 and δ=3. The proof proceeds through explicit calculations using Laplace transforms of squared Bessel bridges and the family of distributions μ_α. In the second part, the paper uses the δ=2 IbPF to construct, via Dirichlet form methods, a weak dynamics for the law of a 2-dimensional Bessel bridge from 0 to 0, claimed to satisfy a regularized version of the conjectured Bessel SPDE. The dynamical half rests on Proposition 4.1, which asserts closability, locality, quasi-regularity, and the isometric identification of the gradient Dirichlet form, but whose proof is deferred to [23], [7], and private communication.","tokens_in":33149,"tokens_out":7448,"duration_ms":76361,"significance":"If the results are correct, the paper provides a valuable extension of the Bessel SPDE programme to general Dirichlet boundary conditions, showing that the renormalised-local-time structure is not an artefact of the zero-boundary setting. The Section 3 IbPFs are derived with substantial explicit computation and are of independent interest. The dynamical construction for δ=2 would be the first weak solution theory for the corresponding Bessel SPDE, complementing the δ=1 case of [7]. However, the dynamical claim is only as strong as the unproved Proposition 4.1; the paper is therefore a solid contribution to the IbPF half but is conditional in the Dirichlet-form half. The manuscript makes no machine-checked claims, but the analytic computations are detailed and reproducible in principle.","major_comments":[{"comment":"Proposition 4.1 asserts that the gradient form (E^2, FC_b^∞(K)) is closable, that its closure is a local quasi-regular Dirichlet form on L^2(ν_2), and that the isometric identity (4.4) holds; no proof is given in the manuscript, which instead refers to Theorem 5.1.3 of [23], Proposition 5.1 of [7], and arguments communicated by R. Zhu and X. Zhu. This proposition is load-bearing: Corollary 4.2, Lemma 4.5, Corollary 4.6, and Theorem 4.7 all rely on the existence and quasi-regularity of this Dirichlet form. Without a complete proof or a fully detailed adaptation of the cited arguments, the dynamical half of the paper is conditional on an unstated external input rather than being established in the manuscript.","section":"Section 4.2, Proposition 4.1"},{"comment":"The proof of Theorem 4.7 is reduced to 'the same arguments as in the proof of Theorem 5.9 in [7]', and Corollary 4.6 is obtained by 'arguing as for the proof of Corollary 5.6 in [7]'. The present setting differs from [7] in the form of the drift approximation, in particular the double limit ε→0, η→0 in (4.7) and the definition of the potential V in (4.9). Moreover, the projection operator Π of Lemma 4.5 inherits the unresolved status of Proposition 4.1. The martingale decomposition and the identification of the zero-energy additive functional are therefore not fully established in the text as it stands; the author should supply the missing details or explicitly verify that all hypotheses of the cited arguments are met in this new setting.","section":"Section 4.4, Theorem 4.7"}],"minor_comments":[{"comment":"In the display immediately preceding equation (3.16), the Taylor-remainder operators appear as T^{-2k}_b and T^{-k}_y, whereas the statement of Theorem 3.1 and the subsequent identification with the distribution μ_{(δ-3)/2} require T^{2k}_b and T^k_y; please correct this sign inconsistency.","section":"Section 3.2, proof of Theorem 3.7"},{"comment":"The symbol μ is used both for the distributions μ_α on R_+ in Section 2.1 and for the probability measure μ_2 on H_2 in Section 4.1; this notation conflict is a potential source of confusion and should be disambiguated.","section":"Sections 2.1 and 4.1"},{"comment":"Equation (1.15) uses ψ_1, ψ_r, and ψ̂_r without defining them; please add a pointer to the definitions in Section 2 or to the corresponding formula in [7].","section":"Introduction, equation (1.15)"},{"comment":"The abstract and introduction state that the results extend to Bessel processes with arbitrary initial conditions and to general Dirichlet boundary conditions, but the dynamical construction in Section 4 is carried out only for the δ=2 Bessel bridge from 0 to 0; please make this scope explicit in the abstract and introduction.","section":"Abstract and Section 1.3.2"}],"recommendation":"major_revision","confidential_remarks":"The decisive issue is the deferred proof of Proposition 4.1, on which the entire dynamical half of the paper depends. If the author can provide a complete proof of Proposition 4.1, or specify a precise published reference containing all its assertions, the paper would be suitable for publication. The reliance on private communication for a load-bearing existence result is not satisfactory for a journal publication. I would encourage the editor to request this revision rather than reject, because the Section 3 IbPF contributions appear technically sound and valuable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the integration by parts formulas are the real product here, and they are in good shape. The dynamical result for δ=2 is genuinely new in aspiration but is currently held together by a proposition the paper does not prove.\n\nWhat the paper actually does: it extends the IbPFs of Elad Altman–Zambotti from zero-to-zero Bessel bridges to arbitrary boundary values and arbitrary initial conditions. Theorem 3.1 covers all δ>0 with separate formulas at δ=1,3, and the proof via pinned bridges and the family of distributions μα is worked out in detail. The conditioning argument that passes from the process case to the bridge case is clean. Section 3 is the substantive, reproducible core; I could not find a gap there, and the reliance on Lemma 3.3 of [7] is legitimate because that lemma is a standard additivity/Girsanov fact, not a disguised version of the target formula.\n\nThe softer half is Section 4. Proposition 4.1, which asserts closability, locality, quasi-regularity and the isometric identification with the gradient form of the two-dimensional Brownian bridge, is stated without proof. The text says it can be proved as in Theorem 5.1.3 of [23] or Prop. 5.1 of [7], and that the arguments were communicated by R. Zhu and X. Zhu. That is transparent but it is still a load-bearing external dependency: Corollary 4.2, the existence of the Markov process, and Theorem 4.7 all rest on it. Lemma 4.5 (projection) is also deferred, and Theorem 4.7's proof refers to the same argument as Theorem 5.9 of [7]. These are omissions of proof, not signs of error. If Proposition 4.1 is accepted, the weak dynamics construction for the two-dimensional Bessel bridge looks plausible; the identification of the drift as the renormalised local time term is exactly the kind of result the field wants. I would not call the dynamical half wrong, but it is conditional in a way the abstract does not fully advertise.\n\nThe paper also does a good job of marking its own limits: the Bessel SPDEs themselves remain conjectural, and the δ=2 transition discussion is honestly speculative.\n\nWho should read it: people working on Bessel SPDEs, Dirichlet form constructions of singular SPDEs, and the interface-model scaling limit program. A serious referee should engage with it, mainly to push for a complete proof of Proposition 4.1 or a clear statement that it is an assumption. I would accept it for peer review, and my report would ask for the Section 4 dependencies to be either proved or explicitly isolated as assumptions.","headline":"The IbPFs for general boundary values are the real contribution and are in good shape; the δ=2 dynamics is conditional on an unproved quasi-regularity proposition.","tokens_in":33693,"tokens_out":2162,"would_cite":true,"duration_ms":22460,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H15","60J60","60J55"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper extends integration-by-parts identities for Bessel bridge laws to arbitrary boundary values, and constructs a Markov process solving a regularized version of the conjectured Bessel SPDE in dimension two.","keywords":["Bessel process","Bessel bridge","integration by parts formula","Dirichlet boundary conditions","renormalised local times","Dirichlet forms","stochastic heat equation","Markov process"],"falsifier":"Compute the energy $E^2(f_n,g_n)$ for smooth cylinder functions whose supports in $K$ are disjoint and concentrate near different points; locality of the closed Dirichlet form requires this energy to vanish in the limit, while non-closability would appear as a bounded-energy sequence with no limit in $L^2(\\nu_2)$. Either failure would directly contradict Proposition 4.1 and invalidate the Markov-process construction of Theorem 4.7.","tokens_in":32650,"feed_emoji":"🎲","tokens_out":15088,"duration_ms":143297,"temperature":0.7,"pith_summary":"This paper asks whether the stochastic PDEs that have Bessel-bridge laws as invariant measures—the so-called Bessel SPDEs—make sense when the bridge is required to start and end at arbitrary nonnegative heights, not just at zero. The author proves that the integration-by-parts formulas that encode the bridges' structure extend to every pair of boundary values $a,a'\\ge 0$ and to Bessel processes with arbitrary starting point, for every dimension $\\delta>0$, with special formulas at $\\delta=1$ and $\\delta=3$. These formulas are the rigorous counterpart of the drift term in the conjectural Bessel SPDEs, expressed through renormalized local times, and they show that the same renormalized structure persists under general Dirichlet boundary conditions. For $\\delta=2$, the paper goes beyond the formal level: using Dirichlet forms, it constructs a Markov process on nonnegative paths whose reversible measure is the law of the two-dimensional Bessel bridge from $0$ to $0$, and proves that this process satisfies a weak, regularized form of the Bessel SPDE. The upshot is that renormalized local times look like an intrinsic feature of Bessel SPDEs below dimension three rather than an artifact of the zero boundary condition.","feed_headline":"Bessel SPDE identities now hold for any endpoints","feed_subtitle":"Integration-by-parts formulas cover arbitrary endpoints and yield a Markov process for the δ=2 Bessel bridge.","key_machinery":"The argument rests on three linked objects. First, the family of measures\n$$\\$Sigma^{{\\delta,r}}$_{a,a'}(dX|b) := \\frac{$p^{{\\delta,r}}$_{a,a'}(b)}{$b^{{\\delta-1}}$} $P^{{\\delta}}$_{a,a'}[dX|X_r=b],$$\nthe candidate Revuz measures of the diffusion local time of the conjectured SPDE solution at level $b$ and time-space point $r$. Second, the analytic family of Schwartz distributions on the half-line, the generalized functions $x_+^{\\alpha-1}/\\Gamma(\\alpha)$, which packages every integration-by-parts formula into a single expression in which the apparent singularity at $\\delta=2$ is cancelled by a vanishing property. Third, for the dynamical result, the Gaussian representation of the $2$-dimensional Bessel bridge: the process $X=\\|\\beta\\|$ with $\\beta$ a two-dimensional Brownian bridge pulls the gradient Dirichlet form on nonnegative paths back to the explicit Gaussian Dirichlet form of the linear heat equation, and convergence of the associated one-potentials identifies the limiting drift.","core_discovery":"The central claim is Theorem 3.1: for every $\\delta\\in(0,\\infty)\\setminus\\{1,3\\}$, every pair $a,a'\\ge 0$, every $\\Phi$ in the linear span of functionals $\\exp(-\\langle m,X^2\\rangle)$, and every $h\\in C_c^2(0,1)$, the law $P^{\\delta}_{a,a'}$ of the $\\delta$-dimensional Bessel bridge satisfies\n$$$E^{{\\delta}}$_{a,a'}(\\partial_h\\Phi(X)) + $E^{{\\delta}}$_{a,a'}(\\langle h'',X\\rangle\\Phi(X)) = -\\kappa(\\delta)\\$int_0^{1}$ h_r \\int_0^\\infty $b^{{\\delta-4}}$ $T^{{2k}}$_b \\$Sigma^{{\\delta,r}}$_{a,a'}(\\Phi(X)|\\cdot)\\,db\\,dr,$$\nwith $\\kappa(\\delta)=(\\delta-1)(\\delta-3)/4$, $k=\\lfloor(3-\\delta)/2\\rfloor$, and analogous formulas at $\\delta=1$ and $\\delta=3$. Here $\\Sigma^{\\delta,r}_{a,a'}(dX|b)$ is the candidate Revuz measure for the local time of a would-be solution at level $b$. For $\\delta=2$, the paper combines this formula with the representation of the $0$-to-$0$ Bessel bridge as the Euclidean norm of a two-dimensional Brownian bridge: it constructs a quasi-regular gradient Dirichlet form, obtains an associated Markov diffusion process, and proves in Theorem 4.7 that this process satisfies the regularized SPDE $\\partial_t u = \\tfrac12\\partial_x^2 u - \\tfrac18\\lim_{\\varepsilon\\to 0}\\lim_{\\eta\\to 0}(\\mathbb{1}_{u\\ge\\varepsilon}/u^3 - \\tfrac{2}{\\varepsilon}\\rho_\\eta(u)/u) + \\xi$ in the sense of additive functionals.","pith_inferences":["Because the integration-by-parts formulas are proven for all boundary values while the Gaussian representation used for the dynamics is only available when one endpoint is zero, the obstacle to a $\\delta=2$ dynamics with $a,a'>0$ is technical rather than structural; approximating general endpoints by pairs with one zero endpoint and passing to the limit in the Dirichlet form is a natural test of w","The limiting drift in Theorem 4.7 is defined through a double limit in $\\varepsilon$ and $\\eta$; whether the limit depends on the mollifier $\\rho$ is not analyzed, and checking that independence numerically on the constructed process would upgrade the weak solution to a more canonical object.","The paper leaves well-posedness of the Bessel SPDEs open; since the integration-by-parts formulas determine the generator on the dense algebra $\\mathcal{S}$, a proof of uniqueness of Markov processes with that generator would turn the weak construction into a full solution of the $\\delta=2$ equation.","The conjectured critical behavior at $\\delta=2$ could be probed by simulating the zero set of the constructed process; because the process is not the norm of the stationary heat equation, such simulations would genuinely test the conjecture rather than reproduce known results."],"forward_implications":["For every $\\delta>0$, the algebraic form of the Bessel-SPDE drift—renormalized local times rather than an explicit convex potential—is the same for arbitrary Dirichlet boundary values as for the zero-to-zero case, so changing the endpoints changes only the boundary condition, not the structure of the equation.","The formulas at $\\delta=3$ and $\\delta>3$ reduce to the previously known integration-by-parts formulas for Bessel bridges, showing the new identities are consistent with the classical regime where the bridge law is a Gibbs measure with explicit potential.","There exists a Markov diffusion process on nonnegative paths of $L^2(0,1)$ with the two-dimensional Bessel bridge law from $0$ to $0$ as its reversible measure, and its martingale part has sharp bracket with Revuz measure $\\|h\\|^2_{L^2}\\nu_2$.","Taken with the earlier $\\delta=1$ construction, this gives weak dynamics for both integer dimensions below 3 that admit a Gaussian representation with zero boundary values.","The paper's conjectures identify $\\delta=2$ as the critical dimension for the number of space points where a solution can simultaneously vanish, by analogy with the known zero-hitting transition for Bessel processes."],"supporting_citations":[{"why":"The prior article on Bessel SPDEs: supplies the zero-to-zero integration-by-parts formulas, the space of exponential functionals, the renormalised-local-time form of the conjectured equations, and the $\\delta=1$ Dirichlet-form construction that this paper extends.","marker":"[7]"},{"why":"A standard monograph on continuous martingales: provides the transition densities of Bessel and squared Bessel processes and the conditioning construction of bridges used throughout Section 2.","marker":"[21]"},{"why":"The article establishing the additivity property of squared Bessel processes: this is the source of the semi-explicit Laplace transforms and time-change identities in Lemma 2.7.","marker":"[22]"},{"why":"A treatise on generalized functions: defines the analytic family of distributions on the half-line used to package all the integration-by-parts formulas in one expression.","marker":"[13]"},{"why":"A thesis cited as the template for proving the closability, locality, and quasi-regularity of the gradient Dirichlet form asserted in Proposition 4.1.","marker":"[23]"},{"why":"Lecture notes on random obstacle problems: provides the classical Bessel-bridge theory, the Gibbs representation for $\\delta\\ge 3$, and the older integration-by-parts formulas recovered in Proposition 3.4.","marker":"[31]"}],"fun_headline_variants":["Bessel SPDEs: arbitrary endpoints, new dynamics","From fixed to free: Bessel SPDEs generalize","Bessel SPDEs now handle any boundary data","Local times extend Bessel SPDEs to all endpoints"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The dynamic half of the paper rests on Proposition 4.1, which asserts that the gradient Dirichlet form for the two-dimensional Bessel-bridge law is closable, local, and quasi-regular so that an associated Markov process exists; the paper does not prove this proposition, saying it follows from a thesis and a prior article with the argument communicated privately, so the existence of the Markov process is an assumed input for Theorem 4.7.","fun_headline_variants_meta":{"raw":{"variants":["Bessel SPDEs: arbitrary endpoints, new dynamics","From fixed to free: Bessel SPDEs generalize","Bessel SPDEs now handle any boundary data","Local times extend Bessel SPDEs to all endpoints"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000504,"raw_usage":{"total_tokens":2513,"prompt_tokens":1047,"completion_tokens":1466,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":663,"completion_tokens_details":{"reasoning_tokens":1400}},"tokens_in":663,"tokens_out":1466,"duration_ms":13688,"temperature":1.0,"reasoning_tokens":1400,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:50:01.176438+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the energy $E^2(f_n,g_n)$ for smooth cylinder functions whose supports in $K$ are disjoint and concentrate near different points; locality of the closed Dirichlet form requires this energy to vanish in the limit, while non-closability would appear as a bounded-energy sequence with no limit in $L^2(\\nu_2)$. Either failure would directly contradict Proposition 4.1 and invalidate the Markov-process construction of Theorem 4.7.","supporting_citations":[{"cited_title":"Elad Altman and L","cited_arxiv_id":null,"evidence_quote":"The prior article on Bessel SPDEs: supplies the zero-to-zero integration-by-parts formulas, the space of exponential functionals, the renormalised-local-time form of the conjectured equations, and the $\\delta=1$ Dirichlet-form construction that this paper extends."},{"cited_title":"Nualart and E","cited_arxiv_id":null,"evidence_quote":"A standard monograph on continuous martingales: provides the transition densities of Bessel and squared Bessel processes and the conditioning construction of bridges used throughout Section 2."},{"cited_title":"Shiga and S","cited_arxiv_id":null,"evidence_quote":"The article establishing the additivity property of squared Bessel processes: this is the source of the semi-explicit Laplace transforms and time-change identities in Lemma 2.7."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"A treatise on generalized functions: defines the analytic family of distributions on the half-line used to package all the integration-by-parts formulas in one expression."},{"cited_title":"Revuz and M","cited_arxiv_id":null,"evidence_quote":"A thesis cited as the template for proving the closability, locality, and quasi-regularity of the gradient Dirichlet form asserted in Proposition 4.1."},{"cited_title":"Zambotti","cited_arxiv_id":null,"evidence_quote":"Lecture notes on random obstacle problems: provides the classical Bessel-bridge theory, the Gibbs representation for $\\delta\\ge 3$, and the older integration-by-parts formulas recovered in Proposition 3.4."}],"review_version":1}